Why are hash functions with minimal collisions not used?

Hash functions with minimal collisions are often either too complex or too slow for practical use. It is important to find a balance between computation speed and hash distribution quality.

Moreover, an ideal collision-free hash function for arbitrary data is impossible due to the limited size of the hash value (e.g., 32 or 64 bits). Therefore, functions that provide sufficiently uniform distribution and are fast to compute are chosen.

In some cases, cryptographic strength of the hash is also important, which influences the choice of the function.

Thus, minimizing collisions is not the only criterion, and often a compromise between speed, simplicity, and distribution is more important.